Aerodynamics B Summary

1. Basic Concepts
1.1 Flow types
If there is friction, thermal conduction or diusion in a ow, it is termed viscous. If none of these things
is present, the ow is inviscid. Inviscid ows do not appear in nature, but some ows are almost inviscid.
A ow in which the density is constant, is termed incompressible. If the density is variable, the ow
is compressible.
The Mach number M is dened as V/a, where V is the airow velocity and a is the speed of sound. If
M < 1, the ow is called subsonic. If M = 1, the ow is called sonic. If M > 1 the ow is called
supersonic.
The ow eld variables p, , T and V = ui +vj +wk represent the ow eld. All these variables are
functions of x, y, z and t (they dier per position and time). However, for a steady ow, the ow eld
variables are constant in time. The ow is steady if
dp
dt
= 0,
d
dt
= 0,
dT
dt
= 0,
du
dt
= 0,
dv
dt
= 0 and
dw
dt
= 0. (1.1.1)
Otherwise the ow is unsteady.
1.2 Gradient, divergence and curl
Consider the vector
=

x
i +

y
j +

z
k. (1.2.1)
The gradient of a scalar eld p(x, y, z) is dened as
p =
p
x
i +
p
y
j +
p
z
k. (1.2.2)
The divergence of a vector eld A(x, y, z) = A
x
i +A
y
j +A
z
k is dened as
A =
A
x
x
+
A
y
y
+
A
z
z
. (1.2.3)
The curl of a vector eld A(x, y, z) = A
x
i +A
y
j +A
z
k is dened as
A =

i j k

z
A
x
A
y
A
z

=
_
A
z
y

A
y
z
_
i +
_
A
x
z

A
z
x
_
j +
_
A
y
x

A
x
y
_
k. (1.2.4)
Note that p gives a vector eld, A gives a scalar eld and A gives a vector eld.
These functions can also be derived for cylindrical coordinates (where p = p(r, , z) and A = A(r, , z))
and for spherical coordinates (where p = p(r, , ) and A = A(r, , )), but those equations do not have
to be known by heart.
1
1.3 Integrals
Given a closed curve C, the line integral is given by
_
C
A ds. (1.3.1)
where the counterclockwise direction around C is considered positive.
Now consider a closed surface S, or a surface S bounded by a closed curve C. The possible surface
integrals that can be taken are
__
S
p dS,
__
S
A dS and
__
S
AdS. (1.3.2)
where dS = ndS with n being the unit normal vector. For closed surface S, n points outward.
Consider a volume . Possible volume integrals are
___

p d and
___

Ad. (1.3.3)
1.4 Integral Theorems
There are several theorems using the integral described in the previous paragraph. If S is the surface
bounded by the closed curve C, Stokes theorem states that
_
C
A ds =
__
S
(A) dS. (1.4.1)
If is the volume closed by the closed surface S, Gauss divergence theorem states that
__
S
A dS =
___

( A)d. (1.4.2)
Analogous to this equation is the gradient theorem, which states that
__
S
p dS =
___

p d. (1.4.3)
2
2. Navier-Stokes Equations
2.1 Continuity equation
The continuity equation is based on conservation of mass. Lets look at a volume with surface S,
which is xed in space. The mass ow out of this volume B is equal to the decrease of mass inside the
volume C.
The mass ow through a certain area dS is V dS. Since dS points outward, were looking at the mass
owing outward. To nd the total mass owing outward, we just integrate over the surface S, to nd
that
B =
__
S
V dS. (2.1.1)
Now lets nd C. The mass in a small volume d is d. The total mass in the volume can be found
by a triple integral. But were not looking for the total mass, but for the rate of mass decrease. So we
simply take a time derivative of the mass. This gives
C =

t
___

d. (2.1.2)
Note that the minus is there, because were looking for the rate of mass decrease. (Not increase!) Using
B = C we can nd the continuity equation

t
___

d +
__
S
V dS = 0. (2.1.3)
Since the control volume is xed, we can pull

t
within the integral. And by using Gauss divergence
theorem, we can rewrite this to
___

t
d +
___

(V) d =
___

t
+ (V)
_
d = 0. (2.1.4)
Now it may be assumed that, for every small volume d in the volume , the integrand is zero:

t
+ (V) = 0. (2.1.5)
Note that in the case of a steady ow

t
= 0, so also (V) = 0. And if the ow is also incompressible,
then V = 0. The value V occurs relatively often in equations and will be discussed later.
2.2 Momentum equation
The momentum equation is based on the principle Sum of forces = Time rate of change of momen-
tum. Lets look once more at a xed volume in space with boundary surface S. First well examine
the forces acting on it. Then well examine the change in momentum.
Two types of forces can act on our volume . Body forces, such as gravity, and surface forces, such as
pressure and shear stress. First lets look at the body forces. Suppose f represents the net body force
per unit mass exerted on the uid inside . On a small volume d, the body force is f d. So the total
body force is
___

f d. (2.2.1)
3
Now lets examine the surface forces. On a small surface dS acts a pressure p, directed inward. But dS
is directed outward, so the actual force vector caused by the pressure is p dS. The total pressure force
therefore is

__
S
p dS. (2.2.2)
The shear stresses on the volume, caused by viscous forces, may be complicated. So lets just dene
F
viscous
as the sum of all the viscous stresses. This makes the total force acting on our volume
F =
___

f d
__
S
p dS +F
viscous
(2.2.3)
Now lets look at the rate of change of momentum in . This consists of two parts. First, particles leave
, taking momentum with them. From the previous paragraph, we know that the mass ow leaving
through dS is V dS. So the ow of momentum that leaves through dS is (V dS)V. The total
momentum leaving therefore is
__
S
(V dS)V (2.2.4)
Second, unsteady uctuations of ow properties inside can also cause a change in momentum. The
momentum of a small volume d is the mass times the velocity, being ( d)V. The total momentum of
can be obtained by integrating. But we dont want the total momentum, but the time rate of change
of momentum. So just like in the last paragraph, we put

t
in front of it to get

t
___

Vd (2.2.5)
We now have calculated both the sum of the forces, and the change in momentum. Its time to put it all
together in one equation
___

f d
__
S
p dS +F
viscous
=
__
S
(V dS)V+

t
___

Vd (2.2.6)
Just like we did in the previous paragraph, we can use the gradient theorem to bring the entire equation
under one integral. Lets dene
viscous
as the part of F
viscous
acting on a small volume d. If we
simplify the equation and split it up in components, we nd
f
x

p
x
+
xviscous
=
(u)
t
+ (uV), (2.2.7)
f
y

p
y
+
yviscous
=
(v)
t
+ (vV), (2.2.8)
f
z

p
z
+
zviscous
=
(w)
t
+ (wV). (2.2.9)
If the ow is steady (

t
= 0), inviscid (F
viscous
= 0) and if there are no body forces (f = 0), these
equations reduce to

p
x
= (uV) (2.2.10)

p
y
= (vV) (2.2.11)

p
z
= (wV) (2.2.12)
If the ow is incompressible ( is constant), we have four equations (the momentum equation has three
components) and four unknowns, being p, u, v and w. It can be solved. But if is not constant, we need
an additional equation.
4
2.3 Energy equation
The energy equation is based on the principle that energy can be neither created nor destroyed. Lets
once more take a xed volume with boundary surface S. We will be looking at the time rate of change
of energy. But rst we make a few denitions. B
1
is the rate of heat added to . B
2
is the rate of work
done on . B
3
is the rate of change of energy in . So all values are rates of changes and therefore have
unit J/s. Putting it all together gives something similar to the rst law of thermodynamics. The relation
between B
1
, B
2
and B
3
is
B
1
+B
2
= B
3
. (2.3.1)
First lets look at B
1
. The heat can increase by volumetric heating (for example due to radiation). Lets
denote the volumetric rate of heat addition per unit mass be denoted by q[J/kg s]. The heating of a
small volume d is q d.
In addition, if the ow is viscous, heat can be transferred across the surface, for example by thermal
conduction. This is a complicated thing, so lets just denote the rate of heat addition due to viscous
eects by

Q
viscous
. Now we know that B
1
is
B
1
=
___

q d +

Q
viscous
. (2.3.2)
Now lets look at B
2
. The rate of work done on a body is F V. Just like in the previous paragraph,
three forces are acting on a small volume d. Body forces (Fd), pressure forces (p dS) and viscous
forces. Lets denote the contribution of the friction forces to the work done by

W
viscous
. Putting it all
together gives
B
2
=
___

(f V) d
__
S
pV dS +

W
viscous
(2.3.3)
To nd B
3
, we look at the energy in . The internal energy in is denoted by e, while the kinetic energy
per unit mass if
V
2
2
. The total energy per unit mass is simply E = e +
V
2
2
.
The particles leaving through the surface S take energy with them. The mass ow leaving through a
surface dS is still V dS. Multiply this by the energy per unit mass gives E(V dS), being the rate of
energy leaving through dS. To nd the total rate of energy leaving, simply integrate over the surface
S.
In addition, if the ow is unsteady, the energy inside can also change due to transient uctuations. The
energy of a small volume d is E d. The total energy can be obtained by integrating over the volume
. But we dont want the total energy, we want the time rate of change of energy. So, just like in the
last two paragraphs, we use

t
. Now we have enough data to nd B
3
, which is
B
3
=
__
S

_
e +
V
2
2
_
V dS +

t
___

_
e +
V
2
2
_
d. (2.3.4)
Putting everything together gives us the energy equation
___

q d+

Q
viscous
+
___

(f V) d
__
S
pVdS+

W
viscous
=
__
S

_
e +
V
2
2
_
VdS+

t
___

_
e +
V
2
2
_
d.
(2.3.5)
Just like in the previous paragraphs, we can follow steps to remove the triple integral. Doing this results
in
q +(f V) (pV) +

Q

viscous
+

W

viscous
=
_

_
e +
V
2
2
_
V
_
+

t
_

_
e +
V
2
2
__
, (2.3.6)
where

Q

viscous
and

W

viscous
represent the proper forms of the viscous terms after being put inside the
triple integral.
5
If the ow is steady (

t
= 0), inviscid (

Q
viscous
= 0 and

W
viscous
= 0), adiabatic (no heat addition,
q = 0) and without body forces (f = 0), the energy equation reduces to

_
e +
V
2
2
_
V
_
= (pV). (2.3.7)
2.4 Equation of state
Now we have ve equations, but six unknowns, being p, , u, v, w and e. To solve it, we need more
equations. If the gas is perfect, then
e = c
v
T, (2.4.1)
where c
v
is the specic gas constant for constant volume and T is the temperature. But this gives us
yet another unknown variable, being the temperature. To complete the system, we can make use of the
equation of state
p = RT. (2.4.2)
We now have seven unknowns and seven equations, which means the system can be solved.
2.5 Substantial derivative
Suppose we look at a very small point in space (from a stationary reference frame). The density changes
according to

t
. But now lets look at a very small volume in space (from a co-moving reference frame).
The time rate of change of this volume is dened as the substantial derivative
D
Dt
. It can be shown
that this derivative is given by
D
Dt
=

t
+ ( V)
D
Dt
=

t
+ ( V) . (2.5.1)
Of course the can be replaced by other variables. The rst

t
is called the local derivative and the
second part (V ) is called the convective derivative.
The substantial derivative can be used to write the Navier-Stokes equations in a simpler form. To do
that, we make use of a vector relation, which is rather similar to the chain rule, being
(V) = ( V) + () V = V+V . (2.5.2)
Applying this relation and the substantial derivative to the continuity equation (equation 2.1.5) gives
D
Dt
+ V = 0. (2.5.3)
Using the same tricks, the momentum equation (equation 2.2.7 to 2.2.9) can be rewritten as
f
x

p
x
+
xviscous
=
Du
Dt
(2.5.4)
f
y

p
y
+
yviscous
=
Dv
Dt
(2.5.5)
f
z

p
z
+
zviscous
=
Dw
Dt
(2.5.6)
If the ow is steady (

t
= 0) and inviscid (F
viscous
= 0), these equations can be simplied even more.
6
Now lets look at the energy equation (equation 2.3.6). In the same way as the above equations, it can
be rewritten. The outcome is
q +(f V) (pV) +

Q

viscous
+

W

viscous
=
D
_
e +
V
2
2
_
Dt
. (2.5.7)
It is conventional to call the earlier forms of the equations (equations 2.1.5, 2.2.7 to 2.2.9 and 2.3.6) the
conservation form (or sometimes the divergence form), while the equations of this paragraph are
called the non-conservation form. In most cases, there is no particular reason to choose one form over
the other.
2.6 Divergence of velocity
The quantity V occurs frequently in equations. Lets consider an amount of air from a co-moving
reference frame. As the air moves, the volume of can change. We will take a look at that change now.
Lets consider a small bit of surface dS of . This surface moves. The change in volume that this piece
of surface causes is V dS. So the total change in volume per unit time can be found, using an integral
over the surface, giving
D
Dt
=
__
S
V dS =
___

( V)d. (2.6.1)
The latter part is known due to the divergence theorem. Note that we have used the substantial derivative
D
dt
instead of
d
dt
since we are considering a moving volume of air, instead of air passing through a xed
volume in space.
If the volume is small enough, such that V is the same everywhere in , then we can nd that
V =
1

D
Dt
. (2.6.2)
This equation states that V is the time rate of change of the volume of a moving uid element per
unit volume. This sounds complicated, but an example will illustrate this fact. If V = 0.8s
1
, then
the volume will decrease by 80% every second (the minus sign indicates a decrease). If V = 1s
1
,
then the volume will double in size every second (that is, as long V remains 1s
1
).
7
3. Aerodynamics Lines and Equations
3.1 Pathlines, streamlines and streaklines
A pathline is a curve in space traced out by a certain particle in time. A streamline is a line where the
ow is tangential. A streakline is the line formed by all the particles that previously passed through a
certain point. However, for steady ows, pathlines, streamlines and streaklines simply coincide.
Of these three lines, streamlines are the lines most used in aerodynamics. But how can we nd the
equation for a streamline? Since the velocity vector is tangential to the streamline at that point, we know
that
Vds = 0. (3.1.1)
Looking at the components of the vectors also shows that
dx
u
=
dy
v
=
dz
w
. (3.1.2)
Note that for 2-dimensional situations this equation reduces to
v dx = udy. (3.1.3)
3.2 Vorticity
If is the angular velocity of a small volume in space, then the vorticity is dened as
= 2 =
_
w
y

v
z
_
i +
_
u
z

w
x
_
j +
_
v
x

u
y
_
k = V. (3.2.1)
So in a velocity eld, the curl of the velocity is equal to the vorticity. If = 0 at every point in a ow,
the ow is called irrotational. The motion of uid elements is then without rotation - there is only pure
translation. If = 0 for some point, then the ow is called rotational.
Note that for 2-dimensional ows the vorticity is given by =
v
x

u
y
. So if the ow is irrotational, then
v
x
=
u
y
. (3.2.2)
This equation is the condition of irrotationality for two-dimensional ow and will be used quite
frequently.
3.3 Circulation
Consider a closed curve C in a ow eld. The circulation is dened as
=
_
C
V ds. (3.3.1)
By denition the integral along C has counter-clockwise as positive direction, but by denition the
circulation has clockwise as positive direction. Therefore the minus sign is present in the equation.
If the ow is irrotational everywhere in the surface bounded by C, then = 0.
8
3.4 Stream functions
Lets consider a 2-dimensional ow for now. If the velocity distribution of the ow is known, equation
3.1.3 can be integrated to nd the equation for a streamline

(x, y) = c. The function

is called the
stream function. Dierent values of c result in dierent streamlines.
If a stream function is known, then the product V at a certain point in the ow can be found, using
u =

y
, v =

x
. (3.4.1)
Now suppose were dealing with incompressible ows, and thus = constant. Lets dene a new stream
function =

/. (Note that has unit [m
3
/(s m) = m
2
/s].) Then equation 3.4.1 becomes
u =

y
, v =

x
. (3.4.2)
In polar coordinates this becomes
V
r
=
1
r
d
d
, V

=
d
dr
. (3.4.3)
3.5 Velocity potential
For an irrotational ow, it is known that = V = 0. There is also a vector identity, stating that
() = 0. Combining these equations, we see that there is a scalar function such that.
V = . (3.5.1)
The function is called the velocity potential. If the velocity potential is known, then the velocity at
every point can be determined, using
u =

x
, v =

y
, w =

z
. (3.5.2)
In polar coordinates, this is
V
r
=
d
dr
, V

=
1
r
d
d
. (3.5.3)
Since irrotational ows can be described by a velocity potential , such ows are also called potential
ows.
3.6 Stream function versus velocity potential
The stream function and the velocity potential have important similarities and dierences. Keep in
mind that the velocity potential is dened for irrotational ow only, while the stream function can be
used for both rotational and irrotational ows. On the contrary, the velocity potential applies for three-
dimensional ows, while the stream function is dened for two-dimensional ows only.
There is another interesting relation between the stream function and the velocity potential. Suppose we
plot lines for constant values of the stream function = constant. The streamlines do not intersect other
stream lines. Now we also plot lines for constant values of the velocity potential = constant, being
so-called equipotential lines. The equipotential lines do not intersect other equipotential lines either.
However, the streamlines and the equipotential lines do intersect. The peculiar thing is that they always
intersect perpendicular. This, in fact, can be mathematically proven. So streamlines and equipotential
lines are orthogonal.
9
4. Basics of Inviscid Incompressible Flows
4.1 Bernoullis equation
An incompressible ow is a ow where the density is constant. Lets assume were dealing with an
incompressible ow. From the momentum equation and the streamline condition, we can derive that
dp = V dV. (4.1.1)
This equation is called Eulers equation. Since the streamline condition was used in the derivation, it
is only valid along a streamline. Integrating the Euler equation between point 1 and point 2 gives
p
1
+
1
2
V
2
1
= p
2
+
1
2
V
2
2
. (4.1.2)
In other words, p +
1
2
V
2
is constant along a streamline.
An inviscid ow is a ow without friction, thermal conduction or diusion. It can be shown that inviscid
ows are irrotational ows. For irrotational ows p +
1
2
V
2
is constant, even for dierent streamlines.
4.2 Continuity equation
In a low-speed wind tunnel the ow eld variables can be assumed to be a function of x only, so A = A(x),
V = V (x), p = p(x), etcetera. Such a ow is called a quasi-one-dimensional ow. From the continuity
equation can be derived that

1
A
1
V
1
=
2
A
2
V
2
, (4.2.1)
for two points in the tunnel. This applies to both compressible and incompressible ows. If the ow
becomes incompressible, then
1
=
2
. The equation then reduces to A
1
V
1
= A
2
V
2
. If we combine this
with Bernoullis equation, we nd
V
1
=

_
2(p
1
p
2
)

_
A1
A2
1
_. (4.2.2)
4.3 Dynamic pressure
The dynamic pressure is dened as
q =
1
2
V
2
. (4.3.1)
Lets suppose that the velocity at some point 0 is zero (V
0
= 0). If the ow is incompressible, it follows
that
p
1
+
1
2
V
2
1
= p
0
q
1
= p
0
p
1
. (4.3.2)
Note that this follows from Bernoullis equation. If the ow is compressible, Bernoullis equation is not
valid and thus p
0
p
1
= q
1
.
4.4 Pressure coecient
The pressure coecient C
p
is dened as
C
p
=
p p

, (4.4.1)
10
where q

=
1
2

V
2

. The subscript denotes that the values are measured in the free stream, as if
being innitely far away from the examined object. For incompressible ows, C
p
can also be written as
C
p
= 1
_
V
V

_
2
. (4.4.2)
4.5 Laplaces equation
If the ow is incompressible, it follows from the continuity equation that
V = 0. (4.5.1)
If the ow is also inviscid, and thus irrotational, it follows that V = 0. It also implicates that there
is a velocity potential such that V = . Combining this with equation 4.5.1 gives
() =
2
= 0. (4.5.2)
This simple but important relation is called Laplaces equation. It seems that the velocity potential
satises Laplaces equation. But what about the stream function? We can recall from the previous
chapter that
u =

y
, v =

x
. (4.5.3)
We can also remember the irrotationality condition, stating that
v
x

u
y
= 0. Inserting 4.5.3 in this
condition gives

x
2
+

2

y
2
= 0
2
= 0. (4.5.4)
So the stream function also satises Laplaces equation, just like the velocity potential function .
4.6 Applying Laplaces equation
Note that the Laplace equation is a linear partial dierential equation. So if we nd multiple solutions

1
, . . .,
n
for it, then any linear combination = c
1

1
+ . . . + c
n

n
is also a solution. So if we nd
a couple of basic solutions to Laplaces equation, and if we add them up in just the right way, we can
display any inviscid incompressible ow.
But how do we know how to put the independent solutions together? We have to make use of boundary
conditions. First, there are the boundary conditions on velocity at innity, stating that, at
innity,
u =

x
=

y
= V

, v =

y
=

x
= 0. (4.6.1)
There are also the wall boundary conditions. The ow can not penetrate an airfoil. So the velocity
at the airfoil edge is directed tangentially. This can be expressed in many ways. If n is the normal vector
at the airfoil surface, then V n = () n = 0. This is called the ow tangency condition. But since
the airfoil edge is a streamline itself, also
surface
= constant.
If we are dealing with neither or , but rather with u and v themselves, things are dierent. If the
shape of the airfoil is given by y
b
(x), then
dy
b
dx
=
_
v
u
_
surface
. (4.6.2)
With those boundary conditions, we can put the elementary solutions to Laplaces equation together to
represent, for example, the ow over a cylinder or over an airfoil. All that is left now, is to nd those
elementary solutions. That is the subject of the next chapter.
11
5. Elementary Flows
5.1 Uniform ows
A uniform ow (oriented in the positive x-direction) is a ow with velocity components u = V

and
v = 0 everywhere. Such a ow is irrotational. It therefore has a velocity potential , which can be shown
to be
= V

x. (5.1.1)
Also the stream function can be determined to be
= V

y. (5.1.2)
Note that these two functions both satisfy Laplaces equation.
In a 3-dimensional world, the velocity potential is the same, and therefore w = 0.
5.2 Source ows
A source ow is a ow where all the streamlines are straight lines emanating from a central point O,
and where the velocity varies inversely with the distance from O. In formula this is
V
r
=

2r
, V

= 0, (5.2.1)
where is the source stength. Such a ow is incompressible (except at point O itself) and irrotational.
If > 0, we are dealing with a source ow. If < 0, we are looking at a so-called sink ow, where the
velocity vectors point inward.
The velocity potential of the ow can be found using the above velocity relations. The result will be
=

2
lnr. (5.2.2)
Note that this function is not dened for r = 0, since the ow is not incompressible there. Identically,
the stream function can be shown to be
=

2
. (5.2.3)
Now lets look at 3-dimensional sources. 3-Dimensional source ows are similar to 2-dimensional ones.
Lets dene as the volume ow originating from the source. The velocity is now, in spherical coordinates,
V
r
=

4r
2
, V

= 0, V

= 0. (5.2.4)
The stream function is not dened for 3-dimensional situations. The velocity potential is
=

4r
. (5.2.5)
5.3 Doublets
Suppose we have a source of strength at coordinates (
1
2
l, 0) and a source of strength (thus being
a sink) at coordinates (
1
2
l, 0). If l 0, we obtain a ow pattern called a doublet. The strength of the
doublet is dened as = l. So as l 0 also . The velocity potential now is
=

2
cos
r
. (5.3.1)
12
Also, the stream function is
=

2
sin
r
. (5.3.2)
The streamlines of a doublet are therefore given by
= c r =

2c
sin. (5.3.3)
It can mathematically be shown that these are circles with diameter d =

2c
and with their centers
positioned at coordinates (0,
1
2
d).
Now lets look at 3-dimensional doublets. Just like in a 2-dimensional doublet, a 3-dimensional doublet
has a 3-dimensional source and sink at a very small distance from each other. The 3-dimensional doublet
strength is dened as = l. The velocity potential then is
=

4
cos
r
2
. (5.3.4)
5.4 Vortex ows
A vortex ow is a ow in which the stream lines form concentric circles about a given point. Such a
ow is described by
V
r
= 0, V

=

2r
, (5.4.1)
where is the circulation. In this case is also called the strength of the vortex. A positive strength
corresponds to a clockwise vortex, while a counterclockwise vortex indicates a negative strength.
Vortex ow is irrotational everywhere except at r = 0, where the vorticity is innite. The velocity
potential is given by
=

2
. (5.4.2)
Also, the stream function is
=

2
lnr. (5.4.3)
There are no 3-dimensional vortex ows. The only way in which three-dimensional vortex ows can occur
is if multiple 2-dimensional vortex ows are stacked on top of each other. This is then, in fact, still a
2-dimensional problem and can be solved with the above equations.
5.5 Elementary ow overview
Flow type Velocity Velocity potential Stream function
Uniform ow in x-direction
u = V

v = 0
= V

x = V

y
Source/Sink
V
r
=

2r
V

= 0
=

2
lnr =

2

Doublet
V
r
=

2
cos
r
2
V

=

2
sin
r
2
=

2
cos
r
=

2
sin
r
Vortex
V
r
= 0
V

=

2r

2
=

2
lnr
13
6. Basic Applications of Elementary Flows
6.1 Nonlifting ow over a cylinder
If we combine a uniform ow with a doublet, we get the stream function
= V

r sin

2
sin
r
= V

r sin
_
1

2V

r
2
_
= V

r sin
_
1
R
2
r
2
_
, (6.1.1)
where R
2
=

2V
. This is also the stream function for a ow over a cylinder/circle with radius
R =
_

2V

. (6.1.2)
The velocity eld can be found by using the stream function, and is given by
V
r
= V

cos
_
1
R
2
r
2
_
, V

= V

sin
_
1 +
R
2
r
2
_
. (6.1.3)
Note that if r = R, then V
r
= 0, satisfying the wall boundary condition. At the wall also V

= 2V

sin.
This means that the pressure coecient over the cylinder is given by
C
p
= 1
_
V
V

_
2
= 1 4 sin
2
. (6.1.4)
6.2 Nonlifting ow over a sphere
Lets combine a uniform 3-dimensional ow with a 3-dimensional doublet. Lets dene R as
R =
3
_

2V

. (6.2.1)
Using the combined stream function, it can be shown that the velocity eld is given by
V
r
= V

cos
_
1
R
3
r
3
_
, V

= V

sin
_
1 +
R
3
r
3
_
, V

= 0. (6.2.2)
At the wall, the velocity is V

=
3
2
V

sin. This means that the pressure coecient over the sphere is
given by
C
p
= 1
_
V
V

_
2
= 1
9
4
sin
2
. (6.2.3)
6.3 Lifting ow over a cylinder
Lets combine a nonlifting ow over a cylinder with a vortex of strength . This results in a lifting ow
over a cylinder. The resulting stream function is
= (V

r sin)
_
1
R
2
r
2
_
+

2
ln
r
R
. (6.3.1)
From the stream function we can derive the velocity eld, which is given by
V
r
= V

cos
_
1
R
2
r
2
_
, V

= V

sin
_
1 +
R
2
r
2
_

2r
. (6.3.2)
14
To nd the stagnation points, we simply have to set V
r
and V

to 0. If

4VR
1, then the solution is
given by
r = R, = arcsin
_

4V

R
_
. (6.3.3)
However, if

4VR
1, then the solution is given by
r =

4V

_

4V

_
2
R
2
, =
1
2
. (6.3.4)
At the surface of the cylinder (where r = R) is the velocity V = V

. Using this, the pressure coecient

can be calculated. The result is
C
p
= 1
_
2 sin +

2RV

_
2
. (6.3.5)
Using this pressure coecient, the drag coecient can be found to be c
d
= 0. So there is no drag. Also,
the lift coecient is
c
l
=

RV

, (6.3.6)
where R =
1
2
c. Now the lift per unit span L

can be obtained from

L

= c
l
q

c =

RV

1
2

V
2

2R =

. (6.3.7)
This equation is called the Kutta-Joukowski Theorem. It states that the lift per unit span is directly
proportional to the circulation. It also works for shapes other than cylinders. However, for other shapes a
complex distribution of sources and vortices may be necessary, as is the subject of the following paragraph.
6.4 Source Panel Method
The source panel technique is a numerical method to use elementary ows. Lets put a lot of sources
along a curve with source strength per unit length = (s). Such a source distribution is called a source
sheet. Note that can be positive at some points and negative in other points.
Now look at an innitely small part of the source sheet. The source strength of this part is ds. So for
any point P, the contribution of this small source sheet part to the velocity potential is
d =
ds
2
lnr, (6.4.1)
where r is the distance between the source sheet part and point P. The entire velocity potential can be
obtained by integrating, which simply gives
=
_
b
a
ds
2
lnr. (6.4.2)
In the source panel method, usually an airfoil (or an other shape) is split up in a number of small straight
lines for which the velocity potential is separately calculated and the boundary conditions are separately
applied.
15
7. Two-Dimensional Airfoils
7.1 Denitions
There are various ways to describe an airfoil. The NACA-terminology is a well-known standard, which
denes the following airfoil properties. The mean camber line is the line formed by the points halfway
between the upper and lower surfaces of the airfoil. The most forward and rearward points of the airfoil
are the leading edge and the trailing edge, respectively. The straight line connecting the leading and
trailing edges is the chord line.
The length of the chord line is dened as the chord c. The maximum distance between the chord line
and the camber line is called the camber. If the camber is 0, then the airfoil is called symmetric. And
nally, the thickness is the distance between the upper and lower surfaces of the airfoil.
In this chapter we will be looking at 2-dimensional airfoils. Were interested in nding c
l
, the lift
coecient per unit length. At low angles of attack , the value of c
l
varies linearly with . The
lift slope a
0
is the ratio of them, so a
0
=
dc
l
d
.
If gets too high, this relation doesnt hold, since stall will occur. The maximum value of c
l
is denoted
by c
l,max
. This value determines the minimum velocity of an aircraft. The value of when c
l
= 0 is
called the zero-lift angle of attack and is denoted by
L=0
.
7.2 Vortex sheets
In the last chapter we treated the source panel method. We put a lot of sources on a sheet. We can also
put a lot of vortices on a curve s. Lets dene = (s) as the strength of the vortex sheet per unit length
along s. The velocity potential at some point P can then be determined, using
d =
ds
2
=
1
2
_
b
a
ds. (7.2.1)
Here is the angle between point P and the point on the vortex sheet were at that moment looking at.
Also a and b are the begin and the end of the vortex sheet.
The circulation of the vortex sheet can be determined to be
=
_
b
a
ds. (7.2.2)
If the circulation is known, the resulting lift can be calculated using the Kutta-Joukowski theorem
L

. (7.2.3)
7.3 Kutta condition
We can put a vortex sheet on the camber line of an airfoil. We can then use boundary conditions and
numerical computation to nd the vortex strength at every point. But it turns out that there are
multiple solutions. To get one solution, we can use the Kutta condition, which states that the ows
leaves the trailing edge smoothly.
What can we derive from this? For now, lets call the angle of the trailing edge. Also lets call V
1
the
velocity on top of the airfoil at the trailing edge and V
2
the velocity at the bottom of the airfoil at the
same point. If is nite, then it can be shown that V
1
= V
2
= 0. However, if 0 (the trailing edge
16
is cusped), then only V
1
= V
2
. Nevertheless, we can derive the same rule from both situations. Namely,
that the vortex strength at the trailing edge is
(TE) = 0. (7.3.1)
7.4 Thin airfoil theory
Suppose we want to calculate the ow over a very thin airfoil by using a vortex sheet in a free stream
ow. We can put vortices on the camber. But the camber line doesnt dier much from the chord line,
so to keep things simple we place vortices on the chord line.
Since the airfoil is thin, it is by itself a streamline of the ow. So the velocity perpendicular to the camber
line is 0. Lets dene z(x) to be the distance between the mean camber line and the chord line, where x
is the distance from the leading edge. The velocity perpendicular to the camber line, caused by the free
stream ow, at position x, can be shown to be
V
,n
= V

_

dz
dx
_
, (7.4.1)
where is in radians. The velocity perpendicular to the mean camber line, due to the vortices, is
approximately equal to the velocity perpendicular to the chord. It can be shown that this velocity
component on a small part d, with distance x from the airfoil leading edge, is
dw =
() d
2(x )
. (7.4.2)
Integrating along the chord gives the total velocity perpendicular to the chord at position x due to the
vortex sheet, being
w(x) =
1
2
_
c
0
() d
x
. (7.4.3)
We have already derived that the velocity perpendicular to the airfoil is zero. So V
,n
+ w = 0, which
results in
1
2
_
c
0
() d
x
= V

_

dz
dx
_
. (7.4.4)
This is the fundamental equation of thin airfoil theory.
7.5 Vortex distributions of symmetric airfoils
If we have a symmetric airfoil, then there is no camber, so dz/dx = 0 everywhere on the airfoil. This
simplies equation 7.4.4 and we might actually try to solve it now. If we make the change of variable
=
1
2
c(1 cos ) and also dene x =
1
2
c(1 cos
0
), we get
1
2
_

0
() sin d
cos cos
0
= V

. (7.5.1)
This is a complicated integral, but it can be solved. The solution will be
() = 2V

1 + cos
sin
. (7.5.2)
We might want to take a closer look on the change of variable we have made. How can we visualize this
change of variable? Imagine the airfoil being the diameter of a circle. Now imagine we are moving over
the top half of the circle, from the leading edge to the trailing edge. The angle we make with respect to
the center of the airfoil corresponds to the point on the airfoil directly below it, as is shown in gure 1.
17
Figure 1: Clarication of the change of variable.
7.6 Lift coecients of symmetric airfoils
In the last paragraph, we found the vortex strength of a thin symmetric airfoil. Using the vortex strength,
we can nd the circulation, which will turn out to be
= cV

. (7.6.1)
Using the Kutta-Joukowski theorem, we can calculate the lift per unit span on the airfoil, which is
L

= c

V
2

. (7.6.2)
The lift coecient now is
c
l
=
L

1
2

V
2

c
=
c

V
2

1
2

V
2

c
= 2. (7.6.3)
So we know have the important conclusion that for thin symmetric airfoils, the lift slope is a
0
= 2.
7.7 Moment coecients of symmetric airfoils
We can use this theory as well to calculate the moment per unit span exerted on the airfoil around, for
example, the leading edge. Lets call M

the moment per unit span around the leading edge. Moment is
force times distance, so dM

= dL

. The minus sign is there due to sign convention. We know that

the lift per unit span is L

, so we nd that dL

d. We also know that d = ()d.

Combining this all gives
M

LE
=
_
c
0
dL

_
c
0
() d. (7.7.1)
Using the familiar change of variable and integrating gives
M

LE
= q

_
c
2
_
2
2 = c
l
q

_
c
2
_
2
. (7.7.2)
The moment coecient about the leading edge now is
c
m,le
=
M

LE
q

c
2
=
c
l
q

_
c
2
_
2
q

c
2
=
1
4
c
l
. (7.7.3)
The quarter-chord point is the point at distance
1
4
c from the leading edge. Taking sum of the moments
about the quarter-chord point gives the moment coecient about the quarter-chord point
c
m,c/4
= c
m,le
+
1
4
c
l
= 0. (7.7.4)
18
The center of pressure is the point around which there is no moment. So the center of pressure is
equal to the quarter-chord position. The aerodynamic center is the point around which the moment
coecient is independent of . Since c
m,c/4
= 0 for every , the quarter point position is also the
aerodynamic center. So the center of pressure and the aerodynamic center are both located at the
quarter-chord point.
7.8 Vortex distributions of cambered airfoils
For cambered airfoils, it is a lot more dicult to solve equation 7.4.4, since
dz
dx
= 0. Mathematicians have
found the solution to be
() = 2V

_
A
0
1 + cos
sin
+

n=1
A
n
sinn
_
. (7.8.1)
We will not show the derivation, since that will be too complicated. You will just have to accept the
equations.
The values A
n
depend on
dz
dx
and A
0
depends on both
dz
dx
and . In fact, using even more complicated
mathematics, it can be shown that
A
0
=
1

_

0
dz
dx
d
0
, A
n
=
2

_

0
dz
dx
cos n
0
d
0
. (7.8.2)
Note that
dz
dx
is the derivative of z(x), taken at point x. So the value of
dz
dx
depends on x. And x also
depends on
0
, since x =
1
2
c(1 cos ).
7.9 Lift coecients of cambered airfoils
Lets take a loot at the lift coecient of the airfoil. The circulation can be found using
= cV

_
A
0
+

2
A
1
_
. (7.9.1)
The lift per unit span now is
L

V
2

c
_
A
0
+
1
2
A
1
_
. (7.9.2)
The lift coecient can be shown to be
c
l
=
L

1
2

V
2

c
= (2A
0
+A
1
) = 2
_
+
1

_

0
dz
dx
(cos
0
1)d
0
_
. (7.9.3)
We now see that the lift slope is once more a
0
=
dc
l
d
= 2. So camber does not change the lift slope.
However, it does change the zero-lift angle of attack, which will be

L=0
= 2 c
l
=
1

_

0
dz
dx
(cos
0
1)d
0
. (7.9.4)
7.10 Moment coecients of cambered airfoils
Just like we did for symmetric airfoils, we can calculate the moment coecient. The result will be
c
m,le
=

2
_
A
0
+A
1

A
2
2
_
=
c
l
4
+

4
(A
2
A
1
) . (7.10.1)
19
We can once more derive the moment coecient with respect to the quarter-chord point. It will not be
0 this time, but
c
m,c/4
=

4
(A
2
A
1
) . (7.10.2)
The value of c
m,c/4
is independent of , so the quarter-chord point is the aerodynamic center. However,
the moment coecient is not zero, so this point is not the center of pressure. The position of the center
of pressure can be calculated to be
x
cp
=
M

LE
L

=
c
m,le
c
c
l
=
c
4
_
1 +

c
l
(A
1
A
2
)
_
(7.10.3)
7.11 Designing a camber line
We used the camber line (described by
dz
dx
) to nd the coecients A
0
, A
1
, . . .. We can also use the
coecients to nd the camber line. We then have several boundary conditions. Of course z(0) = 0 and
z(c) = 0.
First we need to think of suitable coecients for our design. What these coecients will be depends on
what properties we want to give our airfoil. For example, if we want to have c
m,c/4
= 0, then we should
take A
1
= A
2
. If we have determined our coecients, we can nd our camber line by using
dz
dx
= A
0
+

n=1
A
n
cos n
0
. (7.11.1)
7.12 Design lift coecient
Thin airfoils do have a disadvantage. For most angles of attack, the airow separates at the leading edge
(and reattaches afterward for low velocities). This reduces lift. For one angle of attack, the ow smoothly
attaches to the leading edge. This is the so-called ideal or optimal angle of attack
opt
.
Theoretical calculations can show that this only occurs if the vortex at the leading edge is zero, so

LE
= 0. Combining this fact with equation 7.8.1 gives A
0
= 0. Inserting this in equation 7.8.2 results in

opt
=
1

_

0
dz
dx

0
. (7.12.1)
The lift coecient at the optimal angle of attack is called the design lift coecient. Thanks to equation
7.9.3, we can calculate it, using
(c
l
)
design
= A
1
= 2
_

0
dz
dx
cos
0
d
0
. (7.12.2)
20
8. Three-Dimensional Wings
8.1 Induced drag
So far we have looked at two-dimensional (innite) wings. Now lets look at three-dimensional wings.
Lift is created by a high pressure on the bottom of the wing and a low pressure on top of the wing. At
the wing edges, air tries to go from the bottom to the top of the wing. This causes vortices.
These vortices cause a small velocity component in the downward direction at the wing, called downwash.
So the airfoil sees a dierent ow direction than the free stream ow. Even though is the geometric
angle of attack (with respect to the free stream ow), the eective angle of attack
eff
, which
actually contributes to the lift, is dierent. This is such that

eff
=
i
, (8.1.1)
where
i
is the change of the direction of the air ow close to the airfoil.
i
is called the induced angle
of attack.
But the decrease in lift is only small. The real disadvantage is that the lift factor is tilted backward by
an angle
i
. So part of the lift is pointing in the direction of the free stream ow, so it is actually drag.
This drag is called induced drag.
8.2 Coecients, lift and drag
In the last chapter we have dealt with the lift coecient per unit span c
l
. Now we will deal with the
actual lift coecient C
L
of the entire wing. Identically, the lift per unit span L

becomes the total lift L.

Also c
d
becomes C
D
and D

becomes D. The same goes for moment coecients.

In real life, the drag consists of three parts. There is skin friction drag D
f
, pressure drag D
p
and
induced drag D
i
. The rst two are caused by viscous eect, and together form the prole drag. If C
D,p
is the prole drag coecient, then
C
D,p
=
D
f
+D
p
q

S
. (8.2.1)
The induced drag coecient C
D,i
is
C
D,i
=
D
i
q

S
. (8.2.2)
Together they form the total drag coecient, being
C
D
=
D
f
+D
p
+D
i
q

S
= C
D,p
+C
D,i
. (8.2.3)
8.3 Vortex laments
In the last chapter, we considered 2-dimensional vortices. We can put a lot of them in a three-dimensional
curve, being a so-called vortex lament. A vortex lament has a strength . If we now look at any
part of the curve dl, then the velocity dV at some point P, caused by this part, is
dV =

4
dl r
|r|
3
, (8.3.1)
where r is the vector from the part dl to point P. This important relation is called the Biot-Savart
law.
21
There are a few important rules concerning vortex laments. These are called Helmholtzs vortex
theorems.
The strength of a vortex lament is constant along its entire length.
A vortex lament can not end. It is either a closed curve or it is innitely long.
8.4 Horseshoe vortices
We can model three-dimensional wings using vortex laments. Lets take a wing with wing span b, and
put it in a coordinate system such that the tips are positioned at y =
b
2
and y =
b
2
. Now we can let a
vortex lament run from one tip to the other. But a vortex lament may not end. So from the tip, we
let the laments (both ends) run to innity in the direction of the free stream ow (which is dened as
the positive x-direction). The part of the vortex lament on the wing is called the bound vortex. The
two innite parts are the trailing vortices. The entire vortex lament is called a horseshoe vortex,
since it has the shape of a horseshoe (except for the fact that horses dont have innite feet).
Using the horseshoe vortex, we can already, more or less, model the wing. But if we go to the wing
tips, the induced velocity will go to innity, which isnt what happens in real life. So we need to change
our model. Instead of having one horseshoe vortex, running between
b
2
and
b
2
, we put innitely many,
running between y and y, where 0 y
b
2
. We now have a vortex distribution (y) along the wing
and a vortex sheet with strength d(y) behind the wing.
8.5 Induced angle of attack
Now look at a point on the wing with y-coordinate y
0
. The velocity induced by the semi-innite trailing
vortex at position y can be found using the Biot-Savart law. The result will be
dw =
_
d
dy
_
dy
4(y
0
y)
. (8.5.1)
So if we want to nd the entire induced velocity at point y
0
, we need to integrate along the entire wing,
giving
w(y
0
) =
1
4
_ b
2

b
2
_
d
dy
_
dy
(y
0
y)
. (8.5.2)
Using the induced velocity, we can nd the induced angle of attack to be

i
(y
0
) = tan
1
_
w(y
0
)
V

_

w(y
0
)
V

=
1
4V

_ b
2

b
2
_
d
dy
_
dy
(y
0
y)
. (8.5.3)
Note that w is dened positive upward, but the induced angle of attack was dened to be positive directed
downward. Therefore a minus sign is present. It is also assumed that w is small with respect to V

, so
the small angle approximation can be used.
8.6 Finding the vortex distribution
Now lets derive some more expressions. From the previous chapter, we know that the lift coecient per
unit span at the point y
0
is
c
l
= a
0
(
eff
(y
0
)
L=0
), (8.6.1)
22
where a
0
= 2 for thin wings. But the lift coecient can also be found using
L

=
1
2

V
2

c(y
0
)c
l
=

(y
0
) c
l
=
2(y
0
)
V

c(y
0
)
. (8.6.2)
Combining these equations and solving for
eff
gives

eff
=
(y
0
)
V

c(y
0
)
+
L=0
. (8.6.3)
If we put everything together, the angle of attack can be calculated. The result is
=
eff
+
i
=
(y
0
)
V

c(y
0
)
+
L=0
+
1
4V

_ b
2

b
2
_
d
dy
_
dy
(y
0
y)
. (8.6.4)
This equation is the fundamental equation of Prandtls lifting-line theory. For a wing with a
given design, all values are known except . So this is in fact a dierential equation with which can be
found.
If is found, we can nd the lift distribution using the Kutta-Joukowski theorem (L

(y) =

).
Also the induced drag distribution can be found by using
D

i
(y) = L

sin
i
L

i
=

i
. (8.6.5)
From the lift and drag distribution, the total lift and drag can be found, by integrating over the wing
(from
b
2
to
b
2
).
8.7 Elliptical lift distribution
Suppose we have a wing with a circulation distribution given by
(y) =
0

1
_
2y
b
_
2
, (8.7.1)
where
0
is (per denition) the circulation at y = 0. It can now be shown that the induced velocity and
induced angle of attack are
w =

0
2b

i
=
w
V

=

0
2bV

. (8.7.2)
Now lets dene the aspect ratio as
A =
b
2
S
. (8.7.3)
If we rst express as a function of C
L
, ll it in in equation 8.7.2 and use the denition of the aspect
ratio, then we can derive that

i
=
C
L
A
C
D,i
=
C
2
L
A
. (8.7.4)
So the induced drag only depends on the lift coecient and the aspect ratio. Long slender wings thus
give low induced drag.
23
8.8 General lift distribution
Lets suppose we dont know . If we make the change-of-variable y =
b
2
cos , we can use a lot of
complicated mathematics to transform equation 8.6.4 to
() =
2b
c()
N

1
A
n
sinn +
L=0
() +
N

1
nA
n
sinn
sin
. (8.8.1)
In this equation, the coecients A
1
, . . . , A
n
are the unknown coecients that need to be determined.
If N is higher (so if there are more coecients), the result will be more precise. Do not mix up the
coecients A
i
and the aspect ratio A.
To nd A
1
, . . . , A
N
, you have to apply the equation at N points on the wing. Then you have N equations
and N unknowns, which can be solved. You can take any N points on the wing, except for the tips, since
= 0 at those positions.
We can also derive the lift coecient to be
C
L
= A
1
A. (8.8.2)
If we work things out a lot more, we get an expression for the drag coecient, which appears very familiar.
The result is
C
D,i
=
C
2
L
Ae
. (8.8.3)
The number e is called oswalds factor and is dened as
e = A
2
1
_
N

1
nA
2
n
_
1
=
A
2
1
A
2
1
+ 2A
2
2
+. . . +nA
2
n
. (8.8.4)
It is clear that e 1 (with e = 1 only if A
i
= 0 for i 2). For an elliptical lift distribution e = 1, so this
lift distribution is the distribution with the lowest induced drag. However, to minimize induced drag, it
is often more important to worry about the aspect ratio, then about the lift distribution.
24